East Fishkill, N.Y. - IBM Corp.'s Microelectronics Division has scored another coup in the silicon foundry business, forging a multiyear alliance with graphics-chip maker Nvidia Corp.
IBM will make Nvidia's next-generation GeForce graphics processor line, based on 0.13-micron technology. Production will start this summer at IBM's new 300-mm fab in East Fishkill.
The move signals a strategy change for Nvidia (Santa Clara, Calif.), which has relied on one foundry for its IC production-Taiwan Semiconductor Manufacturing Co. A spokesman said Nvidia is moving toward a dual-foundry strategy to reduce risk.
The deal follows reports that Nvidia was unhappy with TSMC over poor yields with its 0.13-micron chips. Until recently, TSMC was struggling to ramp its 0.13-micron process technology-which affected Broadcom, Nvidia, Qualcomm, Via and other customers, analysts said. Late last year, TSMC said it had solved the yield issues.
Officials at Nvidia downplayed such problems. "We enjoy a great relationship with TSMC," the spokesman said.
"As the world's leading fabless semiconductor company designing sophisticated GPUs [graphics processors], we need partners that can deliver leading-edge manufacturing technology and a broad range of products and services," said Jen-Hsun Huang, president and CEO of Nvidia, in a statement.
In fact, IBM is gaining in foundries at the expense of its Taiwanese rivals, analysts said. In March 2002, IBM and Xilinx disclosed IBM would make Xilinx's FPGAs using its most advanced 130- and 90-nm technologies.
Mark Lapedus is editor of Semiconductor Business News, an EE Times network web site.